Intelligent Computing · 2025 · 13 citations · 24 references
ComputingAutonomous SystemsComputer ScienceArtificial IntelligencePath PlanningRoboticsAutonomous ExplorationExploration-exploitation TradeoffPolicy ImprovementPolicy LearningReinforcement LearningReward DesignAdaptabilityAdaptive ExplorationAutonomous Vehicle Path PlanningCuriosity-driven ExplorationCuriosity-driven LearningDeep Reinforcement LearningExploration In Reinforcement LearningRobustness
In the field of path planning, the efficiency and effectiveness of deep reinforcement learning (DRL) methods are often constrained by the algorithms’ exploration capabilities, particularly in dynamic and nondeterministic environments. This paper introduces a novel DRL optimization approach predicated on an action curiosity mechanism, designed to enhance both performance and efficiency in uncertain settings. By incentivizing agents to explore their surroundings more effectively, the action curiosity module amplifies learning efficiency and curtails training duration. The method’s adaptability and stability in intricate or dynamic scenarios are augmented through a dynamically adjusted reward mechanism. To mitigate the issue of strategy degradation stemming from excessive exploration, we incorporate a cosine annealing strategy that finesses parameter adjustments in real time. Extensive experimentation reveals that our enhanced algorithm outperforms conventional methods markedly in terms of success rate and average reward, among other metrics. These experimental outcomes corroborate the proposed method’s robustness and efficacy, laying a solid groundwork for efficient adaptive autonomous path planning within complex and nondeterministic environments.
24
Human-level control through deep reinforcement learning
Volodymyr Mnih, Koray Kavukcuoglu, David Silver et al. · Nature · 2015 · 30K citations
Advances in neural information processing systems 7
Krzysztof J. Cios, Mark E. Shields · Neurocomputing · 1997 · 22.3K citations
Christopher J. Watkins, Peter Dayan · Machine Learning · 1992 · 9K citations · Full text
Comparing Individual Means in the Analysis of Variance
John W. Tukey · Biometrics · 1949 · 4.1K citations